Incorporating Present-on-Admission Indicators in Medicare Claims to Inform Hospital Quality Measure Risk Adjustment Models

Key Points Question Could present-on-admission indicators enhance risk models used by the Centers for Medicare & Medicaid Services to assess acute myocardial infarction, heart failure, and pneumonia mortality and readmission measures? Findings In this comparative effectiveness study including all Medicare fee-for-service beneficiaries hospitalized for acute myocardial infarction, heart failure, or pneumonia at acute care hospitals, the incorporation of present-on-admission indicators into patient-level and hospital-level 30-day mortality and readmission risk models was associated with modest improvement of discrimination in model performance. Meaning These results suggest that accounting for preexisting conditions in hospital quality outcome data could enhance mortality and readmission measure risk models while incurring no additional burden to health care professionals.


eAppendix 1. POA-Exempt Codes
Valid POA indicators are "Y" (Yes, Present on Admission), "N" (No, Not Present on Admission), "U" (Unknown), and "W" (Clinically undetermined). POA exempt codes are "categories and/or codes for circumstances regarding the healthcare encounter or factors influencing health status that do not represent a current disease or injury or are always present on admission" 1 . Review of the POA exempt list revealed additional instances, such as family history codes, in which POA status would not reflect a patient's health status at the time of admission.
To determine the list of POA codes that would be considered "always POA" and subsequently re-coded to POA = Y, we reviewed the POA-exempt codes list through a multistep process with clinical input from a technical working group. First, we created groupings by code classification. Examples of groups include A00-B00 for "Certain infectious and parasitic diseases" and Q00-Q99 for "Congenital malformations, deformations, and chromosomal abnormalities." One group in particular, Z00-Z99 "Factors influencing health status and contact with health services," contained many clinically diverse codes, which we further delineated by commonality of content or coding purpose. All groupings were made in concert with input from a technical working group.
Next, we made initial recommendations for code groups to include as "always POA" and provided rationale for each recommendation based on coding guidance from the CMS ICD-10-CM Coding Guidelines. 1 The technical working group clinical experts provided clinical rationale for categorizing the POA-exempt codes as "always POA" or "don't count as POA" based on their expertise, the rationale from the ICD-10-CM Coding Guidelines, and our recommendations. We excluded POA-exempt codes from the "always POA" list based on any 2) They could potentially be coded as a complication of care during an index admission or POA, such as exposure to toxic substances or medication overdoses; 3) They provided no relevant information about a patient's health status or reason for admission, such as family history codes or encounter codes, which indicate that a patient has an encounter for a procedure but does not specify that the procedure was performed; or 4) They were not mandatory for reporting. The latter exclusion pertained specifically to S00-T88 Injury, poisoning, and certain other consequences of external causes and V00-Y99 External Causes of Morbidity, which are claims collected for the purposes of injury research.
The majority of ICD-10 codes included in the POA-exempt list were subsequent, sequela, or congenital codes. Based on ICD-10-CM Coding Guidelines, we coded these types of codes as "always POA" because, by definition, subsequent and sequela codes should not be used for conditions acquired during a hospitalization in which the patient is receiving active treatment for that condition. Additional groups of codes were further reviewed by clinical experts to determine whether they should be counted as "always POA" or not.  Table C1 reports the hospital-level results among hospitals with 25 or more admissions.
Mean differences in hospital RSRRs between models with and without POA were very small for all three readmission measures, with the largest mean difference of 0.002% (SD=0.16) for HF readmission. The lower and upper quartiles of differences in RSRRs ranged from -0.066 to 0.057% for AMI, from -0.101 to 0.095% for HF, and from -0.069 to 0.069% for PN.
For the AMI, HF, and PN mortality measures, the differences in RSMRs between models with and without POA indicators were relatively larger and had wider ranges than the differences in RSRRs. The mean differences in RSMRs were negative for AMI (-0.013%) and HF (-0.007%). Specifically, the lower and upper quartiles of differences in RSMRs ranged from -0.212 to 0.179% for AMI, from -0.176 to 0.168% for HF, and from -0.471 to 0.468% for PN.